RIva Helm 2.15.0 deployment failure

AI & Data Science Deep Learning (Training & Inference) Riva
1

Please provide the following information when requesting support.

Hardware - GPU (QUADRO RTX 4000)
Hardware - CPU (amd-epyc-milan datacenter)
Operating System
Riva Version 2.15.0
TLT Version (if relevant)
How to reproduce the issue ? (helm install riva-api)

i had riva 2.11.0 running instance deployed on kuberbnetes cluster form “CoreWeave” but when i try to deploy using the 2.15.0 helm chart , i get 1 triton srever running but not on the 50051 port for a grpc request and 1 pod riva-api remains pending and does not deploy .
The service is not responding .

What is the change in the helm , something related to trition groups? i did not change the default configuration

2

You had it deployed earlier, and it worked?

3

Yes it is deployed and working with Riva 2.11.0

4

hi @hamad ,
We now deploy Triton and Riva as separate pods.
However, can you share teh container logs with us?

Thanks

5

Hi
I tried to reproduce the logs with riva 2.16.0 and it worked .
Successfully deployed .

However, the parakeet 0.6b model doesnt seem to be compatible with the RTX Quadro 4000 i was using with riva 2.11.0

What is the compatibility requirments for the parakeet models

6

Hi @hamad ,
You may check the support matrix for Riva here
https://docs.nvidia.com/deeplearning/riva/user-guide/docs/support-matrix.html#riva-2-16-0

Thanks